Bookmarks for April 25, 2025

  • logiscape/mcp-sdk-php: Model Context Protocol SDK for PHP
    A way to use PHP to extend an LLM and give more context, this could be very useful at work
    Tags: php, llm
  • Netflix Codes: find hidden categories on Netflix
    I'm mostly just interested in the way this feels both madly granualar and weirdly arbitrary. What's a "Christmas children & family films for ages 5 to 7", and where's the line between that and "Goofy christmas children & family films". And what if a christmas film is both British *and* Goofy? And are there really so many "hijacking movies" that they need their own subcategory? Apparently so…
    Tags: data, netflix

Bookmarks for April 22, 2025

Bookmarks for April 16, 2025

Bookmarks for April 13, 2025

Bookmarks for April 12, 2025

  • Internet 3.0 and the Beginning of (Tech) History – Stratechery by Ben Thompson
    This is a solid bit of writing from a few years back that explains a lot about the stages of growth of the internet and how we've arrived at the current moment. I think that four years later, his conclusion that we're headed back to a more open and decentralised internet is optimistic – but thinking about Bluesky, and the idea that independent writing plaforms (or, as we used to call them, blogs) are having a resurgence in response to everything, maybe not totaly unfounded.
  • LLMs can't stop making up software dependencies and sabotaging everything • The Register
    Fascinating new exploit wrinkle, and a very strong argument for not using LLMs to generate entire applications – it'll literally decide to try and pull in dependencies that don't exist, and if the app runner doesn't notice, then a bad actor can simply occupy that space with whatever they're like, and completely co-opt the AI-generated app.
    Tags: lmm, exploit